An ARIMA approach to forecasting electricity price with accuracy improvement by predicted errors

Accurate forecasting electricity price is becoming a crucial issue concerned by market participants either for developing bidding strategies or for making investment decisions. Due to the complicated factors affecting electricity prices, accurate forecasting price turns out to be very difficult and usually cannot be achieved by a single forecasting model. This paper proposes an ARIMA approach to price forecasting with accuracy improvement by predicted errors. Resides a conventional model for price forecasting, models for forecasting residual errors are also established iteratively. ARIMA models for forecasting daily average prices, based on historical data of Californian Power Market, are presented to validate the effectiveness of the proposed methodology. Results show that the method only requires easy-implemented low-order models instead of one complex model, while the accuracy of forecasting is improved significantly. The methodology can also be applied to forecasting market clearing prices and electricity loads.